Roles of literacy and life expectancy in promoting economic well-being across developing countries.

Rahman, Matiur ; Khan, M. Moosa

Abstract

This paper studies the roles of literacy and life expectancy in promoting economic well-being (real GDP per capita) using panel data across 99 developing countries over 1990-2008. The panel unit root tests reveal nonstationarity of each variable with I(l) behavior. The Pedroni panel cointegration tests confirm the presence of a long-run equilibrium relationship among the above variables. The estimates of the error-correction model unveil long-run unidirectional causal flows from literacy and life expectancy to real GDP per capita with strong short-run interactive bidirectional feedbacks underscoring the importance of investment in education and healthcare services in promoting economic well-being across the above developing countries.

Keywords: Literacy, Life Expectancy, Well-Being, Panel Cointegration

I. INTRODUCTION

The traditional neoclassical economic growth models followed by Solow (1956) incorporate capital and labor as variable inputs in the production of output subject to the law of diminishing returns to scale. They ignored the roles of non-economic variables such as human capital and human health variables in economic growth. To keep the economy growing, they depended on infusions of exogenous technological progress. Yet the reality is quite contrary that there are other factors outside the realm of neoclassical growth models that are accountable for maintaining high growth performance in selected developing countries. They are addressed in a new paradigm known as endogenous growth models as developed in the mid-1980s (Romer, 1986). They have enhanced the understanding of the mysteries of high growth performance of East Asian economies.

The growth in real GDP per capita is broadly used as a proxy for economic well-being. This emphasizes only the quantity aspect of prosperity without paying due attention to quality of life proxied by life expectancy while that should also be an important component of economic development. Gaining literacy through schooling helps formation of human capital and improving the quality of workforce augmenting productivity. Productivity increases are expected to be correlated with higher wages. Quality of life is reflected through improvement in life expectancy as a result of enhanced access to adequate nutrition intakes and improved healthcare services. Improvement in human capital and longer longevity of people are conjectured to contribute to larger output.

The growth in real GDP per capita is conditional on the initial level of human capital in addition to the initial level of real GDP per capita (Mankiw, Romer, and Well, 1992). Using the World Bank typology, countries are blocked into four, namely, "High Income", "Upper Middle Income", "Lower Middle Income", and "Low Income". Such classifications are used to study the convergence issue (Barro and Sala-iMartin, 1992). The primary focus of this paper is to explore the influences of literacy and life expectancy on real GDP per capita. Most of the studies on this issue utilized time series data. The use of panel cointegration in this paper provides new perspectives. The remainder of the paper is structured as follows. The next section provides a brief survey of the related literature. Individual section thereafter outlines the empirical methodology, reports results, and offers conclusions in sequence.

II. BRIEF SURVEY OF THE RELATED LITERATURE

Social indicators have been used informally for a long time in economics to assess the state of the nation and programs towards national objectives. Measuring people's quality of life emphasizes human well-being and particularly issues of equity, poverty and gender. Social development indicators are a major challenge for policies aiming to foster sustainable human development that involves improving the social, economic, cultural, political and environmental conditions of a nation to develop the present quality of human life without compromising future generations (Medina, 1996).

The conceptualization of human development and the strategies to foster it have varied through history. During the 1960s, the main concern was the economic growth having interest in the productive value of investment in training and education (Colclough, 1993). The assessment of human development was principally concentrated in the value of human capital (Becker, 1964; Schultz, 1961). In the 1970s, the international concern focused upon poverty alleviation and income distribution (Colclough, 1993). International programs of healthcare and primary schooling targeted the poorest segments of the society. By the end of the 1970s, the focus shifted towards growth concerns and social developments as an interdisciplinary approach (Taylor and Jodine, 1983).

The developmental approach, in general, replaced the efforts of human development of the 1970s with an encouragement for privatization and commitment in support of basic educational and health goals. Meanwhile, the United Nations Development Program (UNDP) was emphasizing the need for placing people at the center of development because "people are the real wealth of nations". The policies of the 1990s focused on poverty alleviation by proving the basic services to the poor. Primary education, health care, family planning, and nutrition and self-employment programs were among the most important services.

Unfortunately, a sound measure of human development is not yet available. The "Human Development Index - HDI" developed by the UNDP has significant conceptual limitations which misjudge the measurement of social development. A new social indicator "Literate Life Expectancy" as developed in Lutz is innovative, simple, and accounts for only two essential elements of social development: literacy and life expectancy (Lutz, 1995). Education and healthcare are the leading factors for social development. Basic education and health are simple measured by the number of people who are literate and by the number of years of personal survival, respectively.

Traditionally, nations strive to achieve a higher real GDP per capita and it erroneously considered the single and most important element to measure their national prosperity. The use of real GDP per capita as an indicator of social development fails to capture the distribution of economic progress. This might produce a misleading picture of a country's social development, insofar as it does not reflect important elements of social prosperity such as education and health. The use of literate life expectancy would be a better proxy for social development.

The accumulation of human capital has gained a central role in the recent growth literature. Lucas (1988) has postulated that human capital is an input in the production process like any other; its accumulation implies capital deepening with an associated period of accelerated growth towards a new steady state growth path of output. Moreover, human capital is necessary for the discovery of new technologies and thus its stock is permanently related to the growth rate of output (Aghion and Howitt, 1998; Nelson and Phelps, 1966; Romer, 1990). Bassanini and Scarpetta (2001) find a significant impact of human capital accumulation on output per capita growth. Although there is strong theoretical support for a key role of human capital in the growth process, empirical evidence is not crystal clear. Card (1999) and Psacharopoulos (1994) find that one additional year of schooling is associated with between 5 and 15 percent higher earning across countries. Also, Jorgenson et al. (1987), and Young (1994, 1997) provide some additional support to a significant growth impact of human capital accumulation. In contrast, Benhabib and Spiegel (1994), Pritchett (1997), and Topel (1999) find that the evolution of human capital over time is not statistically related to output growth.

III. EMPIRICAL METHODOLOGY

Panel data, which has both a cross sectional as well as a longitudinal (time series) component provide a convenient way to study phenomena, where a statistically adequate number of cross-sectional observations may not be obtainable at a given point in time. Thus, the combination of a time series and cross-sections can enhance the quality and quantity of data in ways that would be impossible using only one of these two dimensions (Gujarati, 2003). Our study provides an example of such a situation where incorporating observations on the variables over successive time periods allows us to expand the informational content of the data. Furthermore, since the length of the time series is small compared to the number of cross-sections, the effects of autocorrelation are small if not negligible. Panel data estimation models include the constant coefficient (pooled), the fixed effects and the random effects regression models.

In order to test for the existence of a long-run equilibrium relationship among real GDP per capita (y), life expectancy (x) and literacy rate (z) in a heterogeneous panel consisting of 99 developing countries (Appendix I) over the period 1990-2008, the following model is specified:

[y.sub.it] = [[alpha].sub.i] + [[beta].sub.i] [x.sub.it] + [[beta].sub.j] [z.sub.it] + [[gamma].sub.i][D.sub.it] + [e.sub.it] (1)

where i = 1,..., N and t = 1,..., T

In model (1), [[alpha].sub.i] shows the possibility of country-specific fixed effects and [[beta].sub.i] as well as [[beta].sub.j] allow for heterogeneous cointegrating vectors. And, [[gamma].sub.t] represents time dependent common shocks, captured by common-time dummies ([D.sub.it]), that might simultaneously affect all the 99 developing countries included in the study. Model (1) is estimated by the recently proposed Pedroni (2000, 2001) panel Fully-Modified Ordinary Least Squares cointegration technique (hereafter, Panel FM-OLS), which adjusts for the presence of endogeneity between literacy and life expectancy, and serial correlation in the data. This method is an appropriate technique, especially when there are endogeneous macroeconomic factors that can cause co-movements between the above variables.

Before estimating model (1), it is required that the order of integration of the variables be determined by using panel unit root tests. If all variables are found to be I(1), then by using the Pedroni panel cointegration tests (Pedroni, 1999, 2000, 2001), it will be investigated whether they are cointegrated. These above-mentioned tests and techniques are warranted to make sure that no spurious regression phenomenon exists in the estimation of [[beta].sub.i] and [[beta].sub.j]. In order to test for the presence of a unit root in the panel data series under study, recent panel unit root tests proposed by Im, Pesaran and Shin (1997, 2003), Maddala-Wu (1999), and the Breitung (2000) test are employed. In all these tests, the null hypothesis is non-stationarity (for details, see Breitung, 2000). Im, Pesaran and Shin (1997, 2003) have proposed the following panel unit root test statistic, [t.sub.IPS], which is applicable to heterogeneous cross-sectional panels:

[t.sub.IPS] = [square root of N]([bar.t] - E [[t.sub.i]|[[rho].sub.i] = 0]) / [square root of Var] [t.sub.i]|[[rho].sub.i] = 0] (2)

where, N is the number of countries, [bar.t] is the mean of the computed Augmented Dickey-Fuller (ADF) statistics for individual countries included in the panel, [[rho].sub.i] is the autoregressive root, E[[t.sub.i]|[[rho].sub.i] = 0] and Var [[t.sub.i] | [[rho].sub.i] = 0] denote respectively, the moments of mean and variance tabulated obtained from Monte Carlo simulation and tabulated by Im, Pesaran, and Shin (1997, 2003). The statistic [t.sub.IPS] approaches in probability a standard normal distribution as N and T tend to infinity. The Maddala-Wu panel data unit root test is a much more flexible test and is applicable even to unbalanced panels and it is valid for individual ADF tests with different lag-lengths. The Maddala-Wu test statistic [lambda], which has a chi-square distribution with 2N degrees of freedom under the null hypothesis is expressed as:

[lambda] = -2 [N.summation over (i=1)] [l.sub.n] [P.sub.i] (3)

where, [P.sub.i] refers to the probability values from individual ADF unit root tests for each country in the panel. Breitung (2000) formulates a panel unit root test statistic which corrects for the dramatic loss of power associated with the IPS (Im, Pesaran and Shin) test when individual ADF tests include a trend in the specification. Breitung panel unit root test has greater power than that of IPS test.

In order to determine whether in the panel under study, the series [y.sub.it], [x.sub.it] and [z.sub.it] are cointegrated, Pedroni's panel cointegration tests are conducted (see Pedroni (1999). Following Pedroni (1999, 2001), the null hypothesis of no cointegration against the alternative of cointegration is tested using the seven test statistics, proposed by Pedroni, which consist of four panel and three group test statistics. Each of these panel test statistics under appropriate standardization is distributed asymptotically as a normal distribution and expressed as follows:

[[theta].sub.NT] - [mu][square root of N]/[square root of v] [right arrow] N(0,1) (4)

where, [mu] and v are the mean and variance respectively of the underlying individual series. The values [mu] and v are simulated and provided by Pedroni (2000, 2001) and their numerical values depend upon the presence of a constant, time trend, and the number of regressors in the cointegration regression. The rejection of the null hypothesis of no cointegration requires that the absolute value of the calculated test statistics exceed the critical value using four panel cointegration tests in this paper.

Subsequently, the following error-correction model Engle and Granger (1987) is estimated on the evidence of a cointegrating relationship:

[DELTA][[gamma].sub.it] = [alpha] + [k.summation over (t=1)][beta][DELTA][[gamma].sub.it-i1] + [1.summation over (t=1)] [phi] [DELTA][x.sub.it-1] + [m.summation over (t=1)] [psi][DELTA] [z.sub.it-1] + [pi][[??].sub.t-1] + [u.sub.it] (5)

For long-run convergence and causal relationship, the estimated coefficient ([??]) of the error-correction term ([[??].sub.t-1]) is expected to be negative and statistically significant. The estimated [beta], [phi], and [psi] reveal short-run interactive bidirectional feedback relationships.

The appropriate lag-lengths are determined by the Akaike information criterion (Akaike, 1969). Annual data from 1990 through 2008 are obtained from various issues of the World Development Indicators (CD-ROM). The 99 countries studied in this paper are listed in the Appendix.

IV. RESULTS

The panel unit root test results are reported as follows:

Table 1
Unit Root Tests

Variables   [t.sub.IPS]    Maddala-Wu [lambda]    Breitung

y                 -1.185                38.281       0.096
                 (0.125)               (0.312)     (0.538)

x                 -0.912                35.061      -0.805
                 (0.186)               (0.451)     (0.210)

z                 -0.814                37.146      -0.912
                 (0.206)               (0.462)     (0.201)

                      Order of Integration

[DELTA]y          -4.179                52.061       -4.51
               (0.001) *             (0.025) *   (0.000) *

[DELTA]y          -8.163               110.322       -8.56
               (0.001) *             (0.000) *   (0.000) *

[DELTA]y          -5.145               115.041       -6.36
               (0.001) *             (0.000) *   (0.000) *

Note: Probability values are reported in parentheses. The critical
IPS value at 5% level or significance is 1.64. The critical value of
[x.sup.2] is 34.

Table 1 shows that each variable is nonstationary in levels and reveals I(1) behavior providing justification for the application of the methodology as outlined in the preceding section. The Pedroni panel cointegration test results are reported as follows:

Table 2
Pedroni Panel Cointegration Tests @

Test Statistic    Value

V-Statistic       0.881 *
6-Statistic      -6.619 *
pp-Statistic     -7.591 *
ADF-Statistic    -4.274 *

* Denotes significance at 1% level.

@ Only four tests are reported in lieu of seven.

As the above computed test statistics in Table 2 are significant at 1% level, the null hypothesis of no cointegration is rejected. This inference supports the fact that the variables under study depict long-run equilibrium relationship. Thus, the estimation of the error-correction model is in order.

Finally, the estimates of the error-correction model (5) are reported as follows:

Table 3
Estimates of Error-Correction Model (5)

Variable               Coefficient   t-Statistic   Prob.

C                       80.40116      6.762649     0.0000
[[??]i.sub.t-1]        -0.101800     -4.415521     0.0000
[DELTA][y.sub.it.-1]    0.452172      11.88963     0.0000
[DELTA][y.sub.it.-2]   -0.017927     -0.470894     0.6379
[DELTA][y.sub.it.-3]    0.174127      4.430123     0.0000
[DELTA][z.sub.it]       7.185365      0.548620     0.5834
[DELTA][z.sub.it-1]     4.073940      0.298219     0.7656
[DELTA][z.sub.it-2]    -4.967048     -0.371377     0.7105
[DELTA][z.sub.it-3]     3.166438      0.245369     0.8062
[DELTA][x.sub.it]      -1.359010     -0.145981     0.8840
[DELTA][x.sub.it-1]     79.72767     -1.082573     0.2794
[DELTA][x.sub.it-2]     223.8529      2.807383     0.0051
[DELTA][x.sub.it-3]    -207.5076      2.726456     0.0066

[[bar.R].sup.2] = 0.288, F=24.362, DW=1.862.

In Table 3, the coefficient of the error-correction term ([[??].sub.t-1]) and the associated t-value confirm a strong long-run casual flow from changes in life expectancy at birth and adult literacy rate to changes in real GDP per capita. The sum of the coefficients of the subsequent variables and the F-statistic at 24.362 indicate interactive bidirectional positive feedbacks among the variables in the short run. In other words, the above variables positively reinforce each other in the short run for stronger influences on real GDP per capita in the long run. The numerical value of at 0.288 is quite modest and the DW-value at 1.862 indicates near-absence of serial correlation.

V. CONCLUSIONS

Each variable is nonstationary with I(1) behavior. The Pedroni panel cointegration tests depict cointegration among real GDP per capita, life expectancy and literacy rate. The estimates of the error-correction model confirm a long-run equilibrium relationship among the above variables and a unidirectional causal flow from the explanatory variables to the dependent variable. Furthermore, there are evidences of strong short-run interactive and positive bidirectional feedback effects among the above variables. Thus, the evolving economic growth dynamics should include life expectancy and literacy in explaining the growth process more comprehensively.

For policy implications, all developing countries should increase investment in education and healthcare services to enhance economic well-being even further in the long run. Moreover, inclusions of these two variables in the neoclassical growth models would augment their ability to explain economic growth more precisely. Education would also improve the quality of labor force that plays a major role in the modern economic growth process. In brief, social progress and economic well-being ought to advance in tandem as one feeds into the other.

APPENDIX

(List of Countries)

1) Albania

2) Algeria

3) Argentina

4) Armenia

5) Bahrain

6) Bangladesh

7) Barbados

8) Belize

9) Benin

10) Bolivia

11) Botswana

12) Brazil

13) Bulgaria

14) Burundi

15) Cameroon

16) Cape Verde

17) Central African Republic

18) Chad

19) Chile

20) China

21) Colombia

22) Comoros

23) Congo, Rep.

24) Costa Rica

25) Cote d'Ivoire

26) Cyprus

27) Dominican Republic

28) Ecuador

29) El Salvador

30) Estonia

31) Ethiopia

32) Ghana

33) Greece

34) Guatemala

35) Haiti

36) Honduras

37) Hungary

38) India

39) Indonesia

40) Iran, Islamic Rep.

41) Israel

42) Italy

43) Jamaica

44) Jordan

45) Kenya

46) Lao PDR

47) Latvia

48) Lesotho

49) Liberia

50) Lithuania

51) Macao, China

52) Malawi

53) Malaysia

54) Mali

55) Malta

56) Mauritanian

57) Mauritius

58) Mexico

59) Moldova

60) Mongolia

61) Morocco

62) Mozambique

63) Namibia

64) Nicaragua

65) Niger

66) Nigeria

67) Oman

68) Panama

69) Paraguay

70) Peru

71) Philippines

72) Portugal

73) Puerto Rico

74) Romania

75) Russian Federation

76) Rwanda

77) Senegal

78) Singapore

79) South Africa

80) Spain

81) Sri Lanka

82) Sudan

83) Syrian Arab Rep.

84) Tajikistan

85) Tanzania

86) Thailand

87) Togo

88) Trinidad and Tobago

89) Tunisia

90) Turkey

91) Uganda

92) Ukraine

93) United Arab Emirates

94) Uruguay

95) Uzbekistan

96) Venezuela, RB

97) Yemen, Rep.

98) Zambia

99) Zimbabwe

References

Aghion, P. and P. Howitt. (1998), Endogenous Growth Theory. Cambridge, Mass: The MIT Press.

Akaike, H. (1969), Fitting Autoregression for Prediction. Annals of the Istitute of Statistical Mathematics. 21, 243-47.

Barro, R. J. and X. Sala-i-Martin (1992), Convergence. Journal of Political Economy. 100, 223251.

Bassanini, A. and S. Scarpetta (2001), Does Human Capital Matter for Growth in OECD Countries? Evidence from Pooled mean-group Estimates. Working Paper No. 282, Economics Department, OECD.

Becker, G. S. (1964), Human Capital. New York: Columbia University Press.

Benhabib, J. and M. Spiegel (1994), The Role of Human Capital in Economic Development: Evidence from Aggregate Cross-country Data. Journal of Monetary Economics. 43, 143-174.

Breitung, J. (2000), The Local Power of Some Unit Root Tests for Panel Data, in B. Baltagi (ed.), Nonstationary Panels, Panel Cointegration and Dynamic Panels. Advances in Econometrics. 15, 161-178.

Card, D. (1999), The Causal Effects of Schooling on Earnings. in O. Ashenfelter and D. Card (eds.) Handbook of Labor Economics. Amsterdam, North Holland.

Colclough, C. (1993), Human Development: Towards a more Integrated Framework for Planning in the 1990s. Discussion Paper 323. Brighton, England: Institute of Development Studies.

Engle, R. and Granger, C. W. J. (1987), Co-integration and Error-correction: Representation, Estimation, and Testing. Econometrica. 35, 315-329.

Gujarati, D. (2003), Basic Econometrics. Fourth Edition. McGraw-Hill.

Im, K. S., M. H. Pesaran, and Y. Shin (1997), Testing for Unit Root in Heterogeneous Panels. DAE Working Paper No. 9526. University of Cambridge.

Im, K. S., M. H. Pesaran, and Y. Shin (2003), Testing for Unit Roots in Heterogeneous Panels. Journal of Econometrics. 115, 53-74.

Jorgenson, D. W., F. M. Gollop and B. M. Fraumeni (1987), Productivity and U.S. Economic Growth. Cambridge, MA: Harvard University Press.

Lucas, R. E. (1988), On the Mechanics of Economic Development. Journal of Monetary Economics. 22.

Lutz, W. (1995), Literate Life Expectancy. POPNET 26 (Winter), pp. 1-5, Luxemburg, Austria. International Institute for Applied Systems Analysis.

Maddala, G. and S. Wu (1999), A Comparative Study of Unit Root Tests and a New Simple Test. Oxford Bulletin of Economics and Statistics. 61, 631-652.

Mankiw, G. N., D. Romer and D. N. Well (1992), A Contribution to the Empirics of Economic Growth. Quarterly Journal of Economics. 107, 407-37.

Medina, S. (1996), Implementing a New Indicator of Social Development in Mexico: Literate Life Expectancy (LLE). WP-96-103. International Institute for Applied Systems Analysis, Laxenberg, Austria.

Nelson, R. and E. Phelps (1966), Investment in Humans, Technological Diffusion and Economic Growth. American Economic Review. 56, 69-75.

Pedroni, P. (1999), Critical Values of Cointegration Tests in Heterogeneous Panels with Multiple Regressors. Oxford Bulletin of Economics and Statistics. 61, 653-670.

Pedroni, P. (2000), Fully Modified OLS for Heterogeneous Cointegrated Panels, in Baltagi, B. and C.D. Kao (Eds.), Advances in Econometrics, Nonstationary Panels, Panel Cointegration and Dynamic Panels. New York: Elsevier Science. pp. 93-130.

Pedroni, P. (2001), Purchasing Power Parity Tests in Cointegrated Panels. Review of Economics and Statistics. 83, 727-731.

Pritchett, L. (1997), Where has all the Education Gone? Policy Research Working Paper No. 1581. The World Bank. Washington D.C.

Psacharopoulos, G. (1994), Returns to Investment in Education: A Global Update. World Development. 22, 1325-1343.

Romer, P. M. (1986), Increasing Returns and Long-run Growth. Journal of Political Economy. 94, 1002-1037.

Romer, P. M. (1990), Endogenous Technological Change. Journal of Political Economy. 98, 70-102.

Solow, R. M. (1956), A Contribution to the Theory of Economic Growth. Quarterly Journal of Economics. 70, 65-94.

Schultz, T. W. (1961), Investment in Human Capital. American Economic Review. 51, 1-17.

Taylor, C. L., and D. A. Jodine (1983), World Handbook of Political and Social Indicators. Third Edition. New Haven: Yale University Press.

Topel, R. (1999), Labor Markets and Economic Growth in O. Ashenfelter and D. Card (eds.), Handbook of Labor Economics. Amsterdam, North Holland.

Young, A. (1994), Lessons from the East Asia NIC's: A Contrarian View. European Economic Review. 38, 964-973.

Young, M. E. (1997), Social Indicators: A Bibliography with Annotations. Springfield, VA: NTIS.

MATIUR RAHMAN, MBA Director and JP Morgan Chase Endowed Professor of Finance, McNeese State University, Lake Charles, LA, USA

M. MOOSA KHAN, Associate Professor of Finance, Prairie View A&M University, Prairie View, TX, USA